tjp3/zo-3 is critical for epidermal barrier function in zebrafish embryos

Available online at www.sciencedirect.com

316 (2008) 36–49www.elsevier.com/developmentalbiology

Developmental Biology

Tjp3/zo-3 is critical for epidermal barrier function in zebrafish embryos

Tanja K. Kiener, Inna Selptsova-Friedrich 1, Walter Hunziker ⁎

Epithelial Cell Biology Laboratory, Institute of Molecular and Cell Biology, A*STAR (Agency for Science Technology and Research),Singapore 138673, Singapore

Received for publication 8 August 2007; revised 11 December 2007; accepted 30 December 2007Available online 26 January 2008

Abstract

TJP3/ZO-3 is a scaffolding protein that tethers tight junction integral membrane proteins to the actin cytoskeleton and links the conservedCrumbs polarity complex to tight junctions. The physiological function of TJP3/ZO-3 is not known and mice lacking TJP3/ZO-3 show noapparent phenotype. Here we show that Tjp3/Zo-3 is a component of tight junctions present in the enveloping cell layer of zebrafish embryos.Silencing tjp3/zo-3 using morpholinos leads to edema, loss of blood circulation and tail fin malformations in the embryos. The ultrastructure oftight junctions of the enveloping cell layer is disrupted, without affecting the asymmetric distribution of plasma membrane proteins. Morphantsshow a loss of the epidermal barrier, as assessed by an increased permeability of the enveloping cell layer to low molecular weight tracers and ahigher sensitivity of the embryos to osmotic stress. Subjecting wild-type embryos to osmotic stress mimicks the morphant phenotype, consistentwith the phenotype being a direct consequence of failed osmoregulation. Thus, Tjp3/Zo-3 is critical for barrier function of the enveloping celllayer and osmoregulation in early stages of zebrafish development.© 2008 Elsevier Inc. All rights reserved.

Keywords: Tight junction; Zonula occludens; Enveloping cell layer; Epidermis; Kidney; Diffusion barrier; Osmoregulation

Introduction

Tight junctions (TJs) are important components of epithelialtissues where they are required for barrier function, epithelialcell polarity and signaling events in response to cell–cell con-tact (Anderson et al., 2004; Matter and Balda, 2007; Tsukitaet al., 2001). Structurally, TJs are composed of integral mem-brane proteins such as claudins and occludin, which are tetheredto the actin cytoskeleton via scaffold or adaptor proteins(Gonzalez-Mariscal et al., 2003; Tsukita et al., 2001; Van Itallieand Anderson, 2006). Among the best-characterized scaffoldproteins are three closely related members of the membrane-

Abbreviations: AJ, adherens junction; GUK, guanylate kinase; MO,morpholino oligonucleotide; MAGUK, membrane-associated guanylatekinase-like; PCR, polymerase chain reaction; PDZ, PSD95/Dlg/ZO-1; S.E.M.,standard error of the mean; siRNA, small interfering RNA; SH3, Src homology3; TEM, transmission electron microscopy; TJ, tight junction; TJP, tight junctionprotein; WT, wild-type; ZO, zonula occludens.⁎ Corresponding author.E-mail address: [email protected] (W. Hunziker).

1 Present address: Vario Health Institute, Edith Cowan University, JoondalupCampus, Joondalup WA 6027, Perth, Australia.

0012-1606/$ - see front matter © 2008 Elsevier Inc. All rights reserved.doi:10.1016/j.ydbio.2007.12.047

associated guanylate kinase-like (MAGUK) protein super-family, TJP1/ZO-1, TJP2/ZO-2 and TJP3/ZO-3 (Gonzalez-Mariscal et al., 2000). An increasing number of proteins thatinteract with these adaptors have been described, but the roleof individual members of the TJP/ZO protein family in thefunction of TJs remains unclear (Gonzalez-Mariscal et al.,2000). In addition, most of the functional studies to date havefocused on tissue culture cell lines such as renal MDCK ormammary EpH4 cells and the relevance of these proteins in thephysiology of living organisms is poorly understood.

Analysis of the role of ZO proteins in cell polarity, TJ barrierfunction and signaling pathways using tissue culture cellshave been conflicting, perhaps reflecting cell type differencesand/or the extent of protein depletion, depending on whethergene inactivation was achieved by homologous recombinationor silencing. Only minor effects on the assembly kineticsand functions of TJs were detected following the ablation ofTJP1/ZO-1 by homologous recombination in Eph4 mammaryepithelial cells (Umeda et al., 2004) or RNA interference inMDCK renal epithelial cells (McNeil et al., 2006). Silencing ofZO-2 alone also had no discernable effect on barrier functionor polarity of MDCK (McNeil et al., 2006) and EpH4 (Umeda

mailto:[email protected]

http://dx.doi.org/10.1016/j.�ydbio.2007.12.047

37T.K. Kiener et al. / Developmental Biology 316 (2008) 36–49

et al., 2006) cells, but a recent study described effects on bothfunctions in siRNA treated MDCK cells. ZO-3-null cell linesderived from mice with a targeted inactivation of the ZO-3gene, showed no obvious phenotype (Adachi et al., 2006).These findings, while sometimes conflicting, suggest that thelack of a single ZO protein may be largely dispensable for TJstructure and function. Only recently have the physiologicalroles of these proteins been explored using animal models.Surprisingly given the epithelial cell specific expression ofZO-3, mice carrying an inactivated ZO-3 gene present with noapparent phenotype (Adachi et al., 2006; Xu et al., 2008).

Recent evidence supports the notion that when compared tomammals, teleosts may have a more complex repertoire of TJcomponents, perhaps reflecting the need to maintain osmoticbalance in the aqueous environment in which they live. Infugu, for example, claudin genes have undergone an un-precedented expansion, resulting in the expression of over 55different claudin genes in different tissues or developmentalstages (Loh et al., 2004). This represents almost three times thenumber of claudin genes present in mammals. Furthermore,two distinct tjp1/zo-1 and tjp2/zo-2 genes each are present infugu and zebrafish and expressed in a tissue and develop-mental stage specific manner in zebrafish (Kiener et al., 2007).In the present study we explored the functional role of tjp3/zo-3 during zebrafish development. We show that tjp3/zo-3 is acomponent of TJs found in the EVL of zebrafish embryos.Silencing tjp3/zo-3 expression leads to a defective EVLbarrier, resulting in an increased sensitivity of embryos toosmotic stress.

Materials and methods

Antibodies

Commercial antibodies used for immunohistochemistry or Western blotanalysis and their source are listed in the description of the respective techniquesbelow. A GST fusion protein containing the C-terminal part of the zebrafishTjp3/Zo-3 was purified and used to immunize rabbits (BioGenes, Berlin,Germany). The antisera were affinity purified and used for immunoblotting andimmunolabeling.

Zebrafish embryo culture

Care and breeding of zebrafish was carried out according to internationaland institutional standards. The AB wild-type (WT) zebrafish line from thein house stock was used. WT embryos were collected from multiple pairmatings. All embryos were maintained in egg water at 28 °C. For experimentsusing embryos older than 6 hpf, 0.2 mM phenylthiourea was added to the eggwater to prevent pigmentation and tricaine was used as an anesthetic beforeprocessing. Staging was done according to hours postfertilization (hpf)(Kimmel et al., 1995).

Zebrafish embryo microinjection

The following antisense morpholino oligonucleotides (MOs) were obtainedfrom Gene Tools (Philomath, OR): a translation blocking morpholino 5′-GCTCCCATATCGTCATCTCCTCCAT-3′ (hybridizes to the start codon andthe following 25 bases of tjp3/zo-3 mRNA) and a splice inhibition morpholino5′-ACCTCGCCACTTACTTTCGATAACG-3′ (hybridizes to bases 38 to 48 ofthe first exon and bases 1 to 14 of the first intron of the tjp3/zo-3 transcript).Both MOs target the two putative tjp3/zo-3 splice variants in zebrafish (Kiener

et al., 2007). A standard control MO (5′-CCTCTTACCTCAGTTACAATT-TATA-3′) with no target in zebrafish (www.gene-tools.com) was used as control.MO sequences were blasted against NCBI and Ensemble zebrafish genomeassemblies and no significant similarities to other genomic loci besides zebrafishtjp3/zo-3 were found. Zebrafish embryos were pressure injected at the 1-cellstage with 1.5 nl MO or expression vector, or both. tjp3/zo-3 splice MO wasinjected at different concentrations (2 ng, 4 ng, 6 ng, 12 ng) and the phenotypesassessed by morphological criteria. A reproducible phenotype emerged withthe injection of 6 ng tjp3/zo-3 splice MO and this amount was used insubsequent experiments. The conventional translation blocking morpholino(ATGmo) was injected at 1 ng. As a negative control, the same amounts of thestandard control morpholino were injected. For rescue experiments, 10 pg ofpcDNA3.1 expression vector containing an N-terminal Flag tagged zebrafishor dog TJP3/ZO-3 cDNAwere injected alone or together with 6 ng of the tjp3/zo-3 splice MO.

RT-PCR

Total RNAwas isolated from zebrafish embryos using the RNeasy Mini kit(Qiagen). 1 mg of total RNA was amplified with the 1-step RT-PCR kit fromQiagen using gene-specific primers. Reverse transcription was performed at50 °C for 30 min and amplifications with Taq polymerase were done using35 cycles of denaturation (96 °C for 30 s), annealing (51 °C–57 °C for 30 s)and elongation (72 °C for 2 min). The following gene specific primers wereused: tjp3/zo-3 flanking the first intron; 5′-agaggcgaggtgttcagagtggtggac-3′ and5′-aagatcaatgcgtgcagaaaacagagt-3′.

Immunoblotting

To isolate membrane-associated proteins, zebrafish embryos were homo-genized in imidazole buffer (10 mM imidazole pH 7.4, 4 mM EDTA, 1 mMEGTA, 0.2 mM DTT, 100 mg/ml PMSF, 10 mg/ml CLAP) and centrifuged at4 °C and 8000 rpm for 10 min. The pellet was resuspended in 6 M urea buffer(6 M urea, 10 mM Tris pH 8.0, 1 mM EDTA, 150 mM NaCl, 1 mM DTT,100 mg/ml PMSF, 10 mg/ml CLAP), vortexed, and centrifuged at 4 °C and13,000 rpm for 1 h. The protein concentration of the supernatant enriched inthe membrane-associated proteins was determined using the Bradford assay(Pierce). Proteins were separated by SDS-PAGE using 10% polyacrylamidegels and transferred to PVDF membranes (Amersham). Membranes wereincubated with affinity purified anti-Tjp3/Zo-3 (1:500 dilution) or M2 anti-Flag (Sigma; 1:1000 dilution) in PBS, 0.5% BSA, followed by HRP-conjugated secondary antibodies (BioRad; 1:3000 dilution), and processed forchemiluminescence detection (Supersignal West Pico, Pierce and Amershamfilms).

Immunolabeling

For immunofluorescence detection of proteins on whole mount embryos orcryosections, embryos were fixed in 4% PFA in PBS. For cryosectioning,embryos were embedded in 1% bacto-agar in PBS and cryoprotected byovernight saturation in 30% sucrose. 15 μm sections were cut using a Zeisscryostat and collected on polylysine coated slides (Menzel-Glaesser, Germany).Affinity purified anti-Tjp3/Zo-3 (1:250), anti-ZO-1 (Zymed, 1:500), anti-Flag(Sigma, 1:1000), anti-β-catenin (Transduction Laboratories, 1:1000), anti-E-cadherin (Transduction Laboratories, 1:500), anti-PKCζ (Santa Cruz, 1:500) andanti-Na+/K+-ATPase (a6F, 1:100) antibodies were diluted in PBS, 10% NGS,0.5% Triton-X100. Alexa-488 and/or Alexa-596 goat anti-rabbit or anti-mousesecondary antibodies (Molecular Probes) were diluted 1:1000. Images wereacquired using an Axiocam camera on a Zeiss microscope.

In situ hybridization

Plasmids containing zebrafish tjp3/zo-3 (pBluescript) and Pax2b (pScript)cDNAs were linearized and antisense mRNA was transcribed by T7 RNApolymerase (NEB) in the presence of dioxigenin-UTP (Roche). In situhybridization of whole mount zebrafish was performed essentially as described(Hauptmann and Gerster, 1994).

http://www.gene-tools.com

38 T.K. Kiener et al. / Developmental Biology 316 (2008) 36–49

Electron microscopy

Three control, three tjp3/zo-3 ATGmo, and five tjp3/zo-3 MO injectedembryos were sectioned and stained for electron microscopy. For each embryo,two grids were examined for EVL cell contacts. All contacts between adjacentEVL cells were counted and classified into one of three groups dependingon weather the electron dense plaque between adjacent EVL cells characteristicfor TJ showed normal morphology, blebbing, or was absent. Numbers weresubjected to t-test analysis to establish significant differences.

Biotin permeability assay

At 6 hpf, zebrafish embryos were dechorionized and exposed to 1 mg/ml S-NHS-Biotin (Pierce) in PBS for 30 min at 4 °C to prevent endocytosis. Embryoswere then washed in PBS, 100 mM glycine and immediately fixed in 4% PFAfor 1 hr at RT, embedded in agar and cryoprotected in 30% sucrose beforesectioning. Sections were incubated with Alexa-488 labeled streptavidin,viewed under the microscope and digital photographs were acquired using thesame settings and exposure times for each slide.

Osmotic sensitivity and cadmium toxicity assays

Embryos were dechorionized at 6 hpf and grown in either 1× egg water(0.9 mM/1.8 mosM red sea ocean salt), 20× egg water (18 mM/36 mosM oceansalt), 50× egg water (45 mM/90 mosM ocean salt); 100× egg water (90 mM/180 mosM ocean salt); 10× NaCl (157 mM/314 mosM NaCl in egg water); 20×CaCl2 (43 mM/112 mosM CaCl2 in egg water); or 20× MgSO4 (37 mM/74 mosM MgSO4 in egg water). The osmolarity of each solutions was less thanthe estimated physiological osmolarity of zebrafish (230–300 mosM), except for10× NaCl (314 mosM). Mortality was assessed at 1 dpf. Experiments were doneat least in triplicates and subjected to t-test analysis to establish statisticalsignificance. To correlate the mortality to the tjp3/zo-3 morphant phenotype, the

Fig. 1. Tight junction biogenesis in the early zebrafish embryo. Tjp3/zo-3 whole mounimmunofluorescence (I–L) of 2-cell stage/0.75 h (A, E, I); 8–16-cell stage/1.25 hpf (maternally supplied mRNAwith in situ hybridization signals from the 2-cell stage (Aalready present in the 2-cell stage embryo, where they are localized diffusely in tharrowheads). Tjp3/Zo-3 remains absent from cell–cell contact sites at the 8-cell s(J, arrows). Tjp3/Zo-3 gets localized to cell–cell contacts only at the 64-cell stage (G,(K, arrow). By 6 hpf the enveloping cell layer contains Tjp3/Zo-3 and Tjp1/Zo-1 at

number of surviving morphants was counted after 2 dpf and 2-way analysis ofvariance (ANOVA) was performed to establish statistical significance. To assessthe sensitivity to cadmium, embryos were grown in 0.5 mM or 1 mM CdCl2 inegg water from 8 hpf up to 2 dpf. Dead embryos were counted at 2 dpf. Again,experiments were done at least in triplicates and subjected to 2-way ANOVA toestablish statistical significance.

Reversal of edema

At 2 dpf tjp3/zo-3 morphants were transferred into a hyperosmotic 300 mMsucrose solution and the number of embryos with pericardial edema wasdetermined at 3 dpf. Mortality was determined at 3 dpf and 4 dpf for both controland morphant embryos. Experiments were done at least in triplicates andsubjected to t-tests to establish statistical significance.

Results

Tight junction formation and tjp3/zo-3 expression in earlyzebrafish embryos

Tjp3/zo-3 is a maternally supplied mRNA with a uniformexpression from the 2-cell stage to the time of gastrulation asdetected by in situ hybridization (Figs. 1A–D). Apical junctionsare formed as early as the 8-cell stage, when an accumulation ofTjp1 could be detected at cell–cell contact sites (Fig. 1J).However, Tjp3/Zo-3 protein was incorporated into tightjunctions only at the 64-cell stage, when it was first detectedat cell–cell contact sites in the EVL (Fig. 1G). By the time ofgastrulation, the EVL cells were all surrounded with tight

t in situ hybridization (A–D) and Tjp3/Zo-3 (E–H) and Tjp1/Zo-1 whole-mountB, F, J); 64-cell stage/2 hpf (C, G, K); shield stage/6 hpf (D, H, L). Tjp3/zo-3 is a–C). Expression is uniform up to 60% epiboly (D). Tjp1/Zo-1 and Tjp3/Zo-3 aree cytoplasm but not at cell–cell contact sites (E, anti-Tjp3/Zo-3; I, anti-Tjp1,tage (F, arrowheads) but Tjp1/Zo-1 already accumulates at cell–cell junctionsarrows), by a time when Tjp1/Zo-1 is strongly present in the junctional complexall cell–cell junctions (G, H).


junctions (Figs. 1H, L). During somitogenesis tjp3/zo-3 hybrid-ization signals become localized to the olfactory placodes, oticplacodes, pronephric ducts and the epidermis. At the larvalstage, tjp3/zo-3 is strongly expressed in the pronephric ducts,intestine and stomach, while the epidermis maintains a lowlevel of expression. In the anterior embryo, tjp3/zo-3 gene isexpressed in the nose, ears, branchial arches and midbrain(Kiener et al., 2007).

Repression of tjp3/zo-3 expression results in a specificmorphant phenotype

In order to explore the role of Tjp3/Zo-3 in early develop-ment, tjp3/zo-3 expression was knocked down using twodifferent MOs. The first MO was a standard translation blockingmorpholino (ATGmo) spanning the translation start site of thetjp3/zo-3 mRNA. The second was a splice inhibition morpho-lino (MO) against the first exon/intron boundary of the zebrafishtjp3/zo-3 gene, inhibiting splicing and leading to the translationof the first exon and part of the first intron up to an in framestop codon. As a result, a truncated Tjp3/Zo-3 protein con-taining only the first 16 N-terminal amino acids and nofunctional protein domains was translated. Inhibition of splicingwas monitored by RT-PCR using primers flanking the firstintron, resulting in a longer PCR fragment (210 bp) when com-pared to properly spliced wild-type mRNA (97 bp) (see below).A standard control morpholino (Cmo) without known target inzebrafish served as a control. Zebrafish embryos were injectedat the 1-cell stage with tjp3/zo-3 MO, tjp3/zo-3 ATGmo orCmo, and analyzed by live imaging, RT-PCR, Western blot, andimmunolabeling.

Embryos injected with either the tjp3/zo-3 MO or ATGmopresented essentially the same phenotypes. At 1 dpf, morphantembryos had shorter and curved tails with an expanded bloodisland (Fig. 2A). Several fish showed an overall reduction inbody length and first signs of pericardial edema. By 2 dpf,pericardial edema was pronounced and many fish lacked bloodcirculation. Morphants had curved tails and the size of the tailfinwas reduced (Fig. 2A). No phenotypes were observed for Cmoinjected embryos.

To confirm the activity of the MOs, RT-PCR analysis wasperformed on embryos injected with either tjp3/zo-3 MO orCmo. In Cmo injected embryos, primers flanking the first exonamplified a single fragment of 97 bp, indicating proper splicingof exon 1 and 2 (Fig. 2B). In embryos injected with tjp3/zo-3MO and selected for the absence of a phenotype, an additional210 bp fragment was amplified. The longer fragment containedsequence from the first intron, indicating that the MO interferedwith the proper splicing of exon 1 and exon 2. The 210 bp frag-ment was predominantly amplified from MO injected zebrafishselected for the presence of a phenotype.

Western blot analysis and immunolabeling confirmed theloss of Tjp3/Zo-3 in morphants. Protein extracts from 6 hpfembryos showed a loss of Tjp3/Zo-3 even before any phenotypecould be observed (Fig. 2C, lanes 1–3). Abundant Tjp3/Zo-3protein was detected by Western blots in extracts from 2 dpfembryos injected with Cmo, whereas little if any was present in

tjp3/zo-3 MO or ATGmo injected embryos presenting aphenotype. The loss of Tjp3/Zo-3 correlates to the morphantphenotype as MO injected embryos without apparent pheno-type retained Tjp3/Zo-3 (Fig. 2C, lanes 4–7). Furthermore,Tjp3/Zo-3 protein detected by immunolabeling in the olfactoryplacodes, otic placodes, epidermis, pronephric ducts, andintestine of normal or Cmo injected embryos between 1 dpfand 6 dpf, was lost from morphants (Fig. 3). As observed above,the morphant phenotype correlated with the loss of Tjp3/Zo-3 asmorpholino injected embryos selected for the absence of anapparent phenotype still expressed Tjp3/Zo-3 protein in thesestructures (data not shown).

Characterization of morphant phenotypes

To further explore the specificity of the morpholino inducedphenotypes, the tjp3/zo-3morpholinos were injected at differentconcentrations and the phenotypes assessed based on morpho-logical criteria. Injection of MO and ATGmo both caused thesame distribution of phenotypes. The severity of the phenotypeswas dosage dependent, but a 10 fold lower concentration ofATGmo was sufficient to induce the same phenotypes as com-pared to MO. This suggests that blocking ribosome attachmentrather than splice site recognition is more efficient in inhibitingtranslation.

To characterize the different phenotypes, zebrafish embryoswere injected at the 1-cell stage with 6 ng Cmo, 1 ng tjp3/zo-3ATGmo, or 2 ng, 6 ng or 12 ng tjp3/zo-3 MO. Embryonicdevelopment was observed at different time points and pheno-types were classified and counted at the onset of gastrulation(shield stage, 6 hpf), during gastrulation (12 hpf), at theend of somitogenesis (1 dpf), and after organogenesis (2 dpf)(Fig. 4A).

At the shield stage, all injections resulted in ∼15% deathsand b5% dumbbell-shaped embryos. These effects were alsoobserved for Cmo injected embryos and considered the result ofthe injection procedure. The earliest defects could bee seenduring somitogenesis (12 hpf) in a small number (b10%) ofembryos that were slightly elongated with the tailbud pointingaway from the yolk instead of growing towards the vegetal pole.The axis of these embryos was wider than that of WT embryos(data not shown).

At 1 dpf, the mortality rate was ∼20% for Cmo and 2 ng or6 ng MO injected embryos, but ∼30% for 12 ng MO and 1 ngATGmo injected embryos. Of the survivors ∼15% (2 ng) to∼70% (12 ng) had tail malformations, predominantly a curvedtail and/or smaller tail fin. In severe cases, the tail was bent at a45° angle, posterior to the yolk extension (“curved tail” pheno-type). The milder form (“small tailfin” phenotype) presentedwith a tail fin of only approximately half the normal width ofWT embryos (Table 1). After 2 days of development, mortalitylevels further increased in the 6 ng and 12 ng MO groups to∼30% and ∼40%, respectively. Of the survivors, ∼40% (2 ngMO) to ∼80% (12 ng MO) had circulation defects and/or tailmalformations (Table 1; Fig. 4A).

The phenotypes were classified into four groups (I–IV)according to severity. Group I had no circulation with edema

Fig. 2. Characterization of the tjp3/zo-3 morpholino phenotype. Zebrafish embryos were injected with a control morpholino (Cmo), a splice site (MO) or a translationstart morpholino (ATGmo) against tjp3/zo-3. (A) Both tjp3/zo-3 splice and translation start morpholino injected embryos present the same phenotypes. At 1 dpfembryos had shorter and curved tails with an expanded blood island (b, arrowhead). Some individuals showed an overall reduction in body length and the beginning ofpericardial edema (c, arrow). By 2 dpf, pericardial edema was pronounced (e, f, arrows) and many individuals had no blood circulation. Morphants had curved tails andthe size of the tailfin was reduced (e, f, arrowheads). (B) tjp3/zo-3 splice morpholino inhibits tjp3/zo-3 mRNA splicing. RT-PCR analysis was performed on embryosinjected either with control morpholino (Cmo), or tjp3/zo-3 splice morpholino (MO) at 1 dpf. In control morpholino injected embryos, only one product of 97 bpcorresponding to the properly spliced exon1/2 sequence was amplified (lane 3). In embryos injected with tjp3/zo-3 splice morpholino and selected for normalphenotypes, a 210 bp fragment containing the first intron sequence was amplified in addition to the 97 bp WT fragment (lane 2). Abnormal phenotypes werecharacterized by the absence of the 97 bp WT fragment and the presence of the 210 bp exon1/intron1/exon2 fragment (lane 1). (C) The tjp3/zo-3 splice and translationblocking morpholino reduce Tjp3/Zo-3 protein levels. Protein from whole 6 hpf (lanes 1–3) or 2 dpf (lanes 4–8) embryos was extracted and blotted with anti-zebrafishTjp3/Zo-3 antibody. At 6 hpf, low levels of Tjp3/Zo-3 protein were detected in embryos injected with a control morpholino (Cmo, lane 1). No Tjp3/Zo-3 protein couldbe detected in embryos injected with tjp3/zo-3 splice morpholino (MO, lane 2) or tjp3/zo-3 translation blocking morpholino (ATGmo, lane 3). At 2 dpf both splicemorpholino and ATG morpholino injected embryos were selected for morphant phenotypes. These individuals showed a strong reduction of Tjp3/Zo-3 protein levels(lanes 5 and 6) while splice morpholino injected embryos without phenotype had normal levels of Tjp3/Zo-3 protein (lane 7). WT embryos injected with Tjp3/Zo-3expression vector served as a positive control (lane 8).


and a curved tail, category II edema with or without curved tail,group III a curved tail only, and class IVa smaller tailfin only. Athigh doses for both tjp3/zo-3 morpholinos the major phenotypeof the survivors was absent circulation with edema and a curvedtail (2-way ANOVA, pb0.01), at low doses a smaller tailfin(Fig. 4B). Less than 3% cases of edema were observed at lowdoses. Pericardial edema was reversible and embryos of pheno-type categories II, III and IV could be grown to adulthood. Areproducible distribution of phenotypes was observed with 1 ngof the translation blocking morpholino and 6 ng of the tjp3/zo-3splice morpholino and these amounts were used in subsequentexperiments.

The morphant phenotype can be rescued by zebrafish tjp3/zo-3

If the phenotypes described above arise from a loss of tjp3/zo-3, they should be rescued by the tjp3/zo-3 cDNA. 6 ngtjp3/zo-3 splice morpholino was co-injected with 10 pgpcDNA3.1 expression vector containing either a Flag-taggedzebrafish tjp3/zo-3 cDNA or a truncated form lacking theC-terminal Thr–Glu–Leu PDZ binding motif, or a Flag-taggeddog TJP3/ZO-3 cDNA. At 1 dpf, embryos were classified asdead if they consisted of a mass of apoptotic tissue or morphantsif they presented curved tails and an expanded blood island (seeFig. 2A). Embryos that had been co-injected with tjp3/zo-3

Fig. 3. The tjp3/zo-3 morpholino causes the loss of Tjp3/Zo-3 protein expression. Immunofluorescence of whole mount zebrafish at 1 dpf (A–D), 2 dpf (E–F), andcryosections at 6 dpf (G–H) stained with anti-Tjp3/Zo-3. Tjp3/Zo-3 protein is lost in morpholino-injected embryos from 1 dpf up to 6 dpf. The morphant phenotypecorrelates to the loss of Tjp3/Zo-3 since embryos selected for the lack of an apparent phenotype still express Tjp3/Zo-3 protein (data not shown). At 1 dpf there is a lossof Tjp3/Zo-3 labeling in the olfactory placodes (A, B, arrows), otic placodes (A, B arrowheads), and skin (C–F arrowheads). At later stages, Tjp3/Zo-3 protein isabsent from the apical membranes of the pronephric ducts (E–H, arrows), and intestine (G, H, arrowheads).

Fig. 4. The tjp3/zo-3 morpholino effect is specific. Analysis of tjp3/zo-3 morpholino injected embryos at 2 dpf showed that the severity of the phenotype is dosagedependent and that the injection of splice-site morpholino and translation morpholino cause the same distribution of phenotypes. (A) Zebrafish embryos were injectedat the 1-cell stage with 6 ng/1.5 nl control morpholino, 1.5 nl tjp3/zo-3 splice morpholino (MO) at different concentrations (2 ng, 6 ng, 12 ng) and 1 ng/1.5 nl tjp3/zo-3translation blocking morpholino (ATGmo). At 2 dpf the phenotypes were assessed by morphological criteria. The death rate was between 20% (2 ng MO) and 40%(12 ng MO). The phenotypes were classified into 4 groups according to severity: no circulation (with edema and curved tail), edema (with or without curved tail),curved tail only, and reduced tailfin only. At a high dose of tjp3/zo-3morpholino (12 ng) the major phenotype of the survivors was absent circulation with edema and acurved tail (55%), whereas at a low dose (2 ng) it was a reduction of the tailfin (20%) with less than 3% of edema. The injection of only 1 ng of translation blockingmorpholino (ATGmo) resulted in a similar distribution of phenotypes as that of 12 ng splice morpholino (MO). Mean and S.E.M.; n≥100 per experiment; 3–5independent experiments. 2-way ANOVA; pb0.05 (12 ng MO), pb0.01 (1 ng ATGmo). Asterisks indicate statistical significance. (B) Phenotypes of injected embryosat 2 dpf. Control morpholino injected embryo (a); tjp3/zo-3 splice morpholino injected embryos (b, d); tjp3/zo-3 translation blocking morpholino injected embryos(c, e). Mild phenotypes presented with crooked tail buds and reduced tailfin size (b, c, arrowheads), while severely affected embryos had no circulation and often anaccumulation of red blood cells (d, e, arrows).



splice morpholino and tjp3/zo-3 cDNA and had no apparentphenotypes were considered “rescued”. Approximately 10%wild-type embryos injected with the Flag-tagged zebrafish butnot the dog cDNA for tjp3/zo-3 presented with a dramaticallyshortened, upwards curved body axis and apoptosis in the brainby 1 dpf, likely due to overexpression of the protein.

Co-injection of tjp3/zo-3 splice morpholino and the full-length or truncated Flag-tagged zebrafish tjp3/zo-3 cDNA lead

Fig. 5. Rescue of the phenotype. Tjp3/zo-3 morphants can be rescued by full-lengthPDZ-binding motive, but not dog TJP3/ZO-3. (A) tjp3/zo-3morphants can be rescuedcontaining full-length Flag-tagged zebrafish tjp3/zo-3 (zf tjp3) or mutant zebrafish tjdog TJP3/ZO-3 cannot rescue the morphant phenotype, even tough the protein is expnot shown). Phenotypes were counted at 1 dpf. Mean and S.E.M.; n≥100 per experimstatistical significance. (B) Rescued embryos showed a straight body axis, little (arroembryos presented with the morphant phenotype as described in Figs. 2 and 4. (C) Imwith pcDNA3.1 Flag-tagged zebrafish tjp3/zo-3 (c, d) at 2 dpf. WT embryos were labEctopically expressed Tjp3/Zo-3 localizes accurately. Anti-Flag staining was detecte(D) Anti-Flag Western blot at 2 dpf. Embryos were injected with control morpholinoand pcDNA3.1 zebrafish tjp3/zo-3 (rescue). Ectopic Flag-tagged Tjp3/Zo-3 is expreembryos (lane 2). No Flag labeling was detected in controls (lane 3, different experimcheck for the activity of the splice morpholino in rescued embryos. In control embryproperly spliced 97 bp wt band was amplified predominantly. However, tjp3/zo-3morescued embryos (lane 4).

to a significant reduction in the number of morphants (2-wayANOVA, pb0.01) with a corresponding increase in the numberof apparently normal embryos, showing that the morpholinoinduced phenotype can be rescued. Interestingly, however, thephenotype could not be rescued by co-injection of the dog Flag-tagged TJP3/ZO-3 cDNA (Fig. 5A). Fractionation of zebrafishembryo homogenates into cytosol and membrane fractionsshowed that the bulk of the dog TJP3/ZO-3 was present in the

zebrafish tjp3/zo-3 and a truncated zebrafish tjp3/zo-3 without the C-terminalby co-injection of tjp3/zo-3 splice morpholino and pcDNA3.1 expression vectorp3/zo-3 without the PDZ-binding motif (zf tjp3ΔP). Co-injection of pcDNA3.1ressed and enriched in membrane associated protein fractions of zebrafish (dataent; 2–6 independent experiments. 2-way ANOVA; pb0.01. Asterisks indicatewhead) or no edema, and only mild tailfin malformations (arrow). Non-rescuedmunofluorescence on cryosections of WT embryos (a, b) and embryos injected

eled with Tjp3/Zo-3 antibody to detect endogenous Tjp3/Zo-3 localization (a, b).d in the apical side of the pronephric ducts (c, arrowheads) and in the skin (d).(Cmo), pcDNA3.1 zebrafish tjp3/zo-3 (zf tjp3), or co-injected with morpholinossed in embryos injected with tjp3/zo-3 alone (lane 1) as well as in the rescuedent). (E) RT-PCR analysis was performed using total RNA of 2 dpf embryos to

os (lane 1) and in morpholino injected embryos without phenotypes (lane 3), therphants had a strong unspliced band at 210 bp (lane 2), which was also present in


membrane-associated fraction (data not shown), consistent withits incorporation into TJs.

Rescued embryos were nearly identical to wt embryos. Theyhad a straight body axis, little or no edema, and only occa-sionally could we observe slight tailfin malformations. The 15%to 20% of non-rescued embryos presented the classical mor-phant phenotype with a shorter and/or curved body axis, severepericardial edema and pronounced tailfin malformations asdescribed above (Fig. 5B).

Immunolabeling with Flag antibodies confirmed that injec-tion of pcDNA3.1 vector leads to the expression of ectopiczebrafish Tjp3/Zo-3. Flag-tagged protein accurately localized tothe lumenal pole of epithelial cells of the pronephric ducts, aswell as to the skin (Fig. 5C).

To confirm expression of the ectopic zebrafish Tjp3/Zo-3 inrescued embryos, protein was extracted from 2 dpf embryosand blotted with anti-Flag antibody. Flag-tagged zebrafish Tjp3/Zo-3 was detected in both embryos injected with zebrafish tjp3/zo-3 alone as well as in rescued embryos where zebrafish tjp3/zo-3 was co-injected with the tjp3/zo-3 morpholino (Fig. 5D,lanes 1–2). No Flag labeling was observed in control embryos(Fig. 5D, lane 3).

Since not all tjp3/zo-3 MO injected embryos displayed aphenotype, it was important to confirm that the splice mor-pholino was indeed active in the rescued embryos. In Cmoinjected embryos and in MO injected embryos selected for theabsence of phenotypes, the properly spliced 97 bp WT RT-PCRfragment was predominantly amplified from 2-dpf embryoRNA (Fig. 5E). Tjp3/zo-3 morphants showed a prominent210 bp fragment, indicative of failed splicing due to the actionof the morpholino. Importantly, the 210 bp fragment was alsopresent in rescued embryos. The slightly lower levels of the210 bp compared to the 97 bp fragment in the rescued embryosmost likely reflects the presence of a small fraction ofindistinguishable “normal” embryos where the splice morpho-lino was inactive. Thus, zebrafish Tjp3/Zo-3 can rescue themorpholino-induced phenotypes.

Tight junction structure in the EVL of tjp3/zo-3 morphants isaltered

In the early zebrafish embryo, the EVL, a specialized layer offlat and elongated cells, forms a primary “skin”, which serves asa barrier between the embryo proper and the surrounding freshwater environment (Kimmel et al., 1990). Given the function ofthe EVL as a barrier and the incorporation of Tjp3/Zo-3 into TJsas early as the 64-cell stage (see Fig. 1G), we speculated that theEVL may establish TJs and that these may be affected by theloss of Tjp3/Zo-3 in the morphants, leading to some of thephenotypes described above.

Immunolabeling of embryos at the onset of gastrulation(6 hpf) showed a concentration of Tjp3/Zo-3 and Tjp1/Zo-1 atsites of cell–cell contact facing the external environment (Fig.6A). At least by real-time RT-PCR, no significant changes in theexpression levels for the other Tjps were observed. Analysis ofembryo sections by transmission electron microscopy (TEM)confirmed the presence of electron dense regions where the

intercellular space between the lateral membranes of adjoiningcells was obliterated, a characteristic feature for TJs (Fig. 6B).In ∼40% of morpholino injected embryos, the Tjp3/Zo-3 pro-tein was lost from the majority of the cell–cell contact sites inthe EVL (Fig. 6A, b) whereas Tjp1 could still be detected atthese sites (Fig. 6A, d).

To determine if the tjp3/zo-3 morpholino affected the archi-tecture of TJs, morphant embryos were analyzed by TEM.Cell–cell contacts between EVL were analyzed for the presenceof TJs and these were classified into three groups based on theirmorphology. The first group included apparently normal TJswith the plasma membrane of neighboring EVL cells in closecontact throughout the length of the adhesion plaque. A secondclass represented TJs with discontinuities in the adjoiningplasma membranes, resulting in blebs. The third group repre-sented the complete loss of the characteristic electron denseplaque.

In control embryos, TJs were readily detected at all contactsites between EVL cells (Fig. 6B, arrow) and only ∼20% of TJsshowed discontinuities. Splice morpholino injected embryosretained TJs between ∼60% of EVL cells, but more than 40%of these TJs showed blebbing (Fig. 6B, red arrows). Theremaining contact sites lacked electron dense plaques (Fig. 6B,arrowheads). When embryos were injected with ATGmo, asmany as 60% of EVL cell–cell contact sites did not have avisible TJ plaque (Fig. 6B, arrowheads). Quantification of theseresults is shown in Fig. 6C. There was a significant reduction ofnormal electron dense TJ plaques in both MO and ATGmoinjected embryos (t-tests, pb0.05). These findings confirm adefect in the morphology of TJs in the EVL cell layer of tjp3/zo-3 morphants.

As injection of both splice and ATG morpholino resulted inthe same phenotypes, e.g. the loss of the protein already at 6 hpfand the disrupted TJ ultrastructure, only the data for the splicemorpholino is shown for subsequent experiments.

The asymmetric distribution of plasma membrane proteins isunaffected in tjp3/zo-3 morphants

TJs have been implicated in maintaining the asymmetricdistribution of proteins to the apical and basolateral plasmamembrane domains of epithelial cells (i.e. “fence function”).We therefore analyzed the polarized distribution of severalmarkers in the EVL at 6 hpf. Since tjp3/zo-3 is also highlyexpressed in the pronephric ducts (Kiener et al., 2007), thedistribution of apical and lateral proteins in renal epithelialcells was also analyzed at 2 dpf. E-cadherin and β-catenin,two markers of adherens junctions, remained restricted tothe lateral domain of the cells of the EVL of morpholino-injected embryos (Fig. 7A). Similarly, despite the loss ofTjp3/Zo-3 in the pronephric ducts of the morphants (Fig. 7B,d), E-cadherin and Na+/K+-ATPase retained their basolateraldistribution (Figs. 7B, d and B, f) and atypical PKCζremained localized to the apical domain (Fig. 7B, f). Wholemount in situ hybridization to detect expression of the earlykidney marker pax2b showed a normal development of thepronephric ducts in morphants (Fig. 7BA, a, b), indicating

Fig. 6. The loss of Tjp3/Zo-3 leads to a disruption of the TJ ultrastructure. (A) Cryosections of 6 hpf embryos stained with anti-Tjp3/Zo-3 (a, b) and anti-Tjp1/Zo-1(c, d). Embryos were injected with control morpholino (a, c, Cmo) or tjp3/zo-3 splice morpholino (b, d, MO). Tjp3/Zo-3 protein is absent from most TJs in the EVL ofmorpholino injected embryos (b), but no phenotype is observed at that stage yet. The same embryos show a slight reduction in Tjp1/Zo-1 staining (d). (B) TEM ofmorpholino injected embryo sections show a marked loss and/or a disruption of the electron dense plaque of junctions in the EVL of morphants at 6 hpf. In controlmorpholino injected embryos, TJs can be observed at every cell–cell contact between neighboring cells of the EVL (Cmo, arrow). In the splice morpholino injectedembryos, more than 40% of the EVL cell–cell contacts had discontinuities (blebbing) (MO, red arrows) and another 40% of contacts did not have any electron denseplaque (MO, arrowheads). Embryos injected with tjp3/zo-3 translation blocking morpholino had an even greater number of missing TJs (ATG, arrowheads).(C) Statistical analysis of TJ integrity as seen in TEM at 6 hpf. Two control and five morpholino injected embryos were analyzed. All contacts between adjacent EVLcells were classified into one of three groups. 1) Normal TJs, where cell membranes were in close contact throughout the length of the TJ, 2) discontinuous TJs withblebs, where cell membranes detached from each other, thus reducing the effective length of the TJ seal, and 3) junctions where no electron dense plaque was foundbetween neighboring EVL cells. About 25% of TJs in control embryos had blebs (bar 1). In splice morpholino injected embryos, 47% of TJs showed blebbing andanother 40% of the junctions were missing an electron dense plaque (bar 2). 62% of EVL cell–cell contacts in ATGmo injected embryos lacked an electron denseplaque (bar 3). N=3 for Cmo, n=3 for ATGmo, n=5 for splice MO; number of cell–cell contacts counted per embryo 6–9; 2 independent experiments; t-test;pb0.05.


that Tjp3/Zo-3 is not critical for renal development. Similarfindings were made for 1 dpf embryos and for ATGmo injectedembryos (Supplemental Data I). Thus, epithelial cell polarity inthe EVL and the pronephric ducts is not affected in tjp3/zo-3morphants.

Tjp3/Zo-3 is required for epidermal barrier function

We next determined weather a second function of TJs, theestablishment of paracellular diffusion barriers, was affected inthe EVL of tjp3/zo-3 morphant embryos.

Fig. 7. The asymmetric distribution of plasma membrane proteins is maintained in tjp3/zo-3 morphants. (A) Epithelial cell polarization in the EVL. Cryosections ofcontrol morpholino (a, c) and tjp3/zo-3morpholino (b, d) injected embryos at 6 hpf. The EVL, a monolayer of epithelial cells that surrounds the blastoderm, is shown.EVL cells retain their elongated, flattened morphology in the morphants (b, d). These cells are still polarized with adherens junctions along their basolateralmembranes, as shown by labeling with anti β-catenin (a, b) and anti pan-Cadherin (c, d) antibodies. Both markers remained restricted to the basolateral surfaces (lateralmembranes, arrowheads; basal membranes, arrows). (B) Epithelial cell polarization in the pronephric ducts. Whole mount in situ hybridization of Pax2b (a, b);Immunofluorescence on cryosections: anti-Tjp3/Zo-3 (green) and anti-E-cadherin (red) (c, d); anti-PKCζ (green) and anti-Na+/K+-ATPase (red) (e, f). Controlmorpholino 1 dpf (a, c, e); tjp3/zo-3 morpholino 1 dpf (b, d, f). In situ hybridization with Pax2b shows that the pronephric ducts extend normally in tjp3/zo-3morphants (a, b, arrowheads). Morphants lost Tjp3/Zo-3 protein from the apical side of the pronephric ducts as shown above (c, d). E-cadherin staining of thebasolateral membranes of the pronephric ducts is maintained in morphants (d, arrows). The polarized distribution of the basolateral Na+/K+-ATPase (e, f, arrows) andthe apical PKCζ (e f, arrowheads) are also maintained in the pronephric duct.


First, we analyzed the diffusion of sulfo-NHS-biotin, a small(MW∼440 Da) membrane impermeable tracer often used toassess the intactness of the TJ barrier (Fesenko et al., 2000;Furuse et al., 2002; Merzdorf et al., 1998). Control morpholinoor tjp3/zo-3 morpholino injected embryos were dechorionizedat 6 hpf and exposed to sulfo-NHS-biotin for 30 min. Followingfixation, the biotin was detected in embryo sections usinglabeled streptavidin. In controls, only a weak staining in theEVL cell layer and the underlying blastoderm was observed(Fig. 8A). In contrast, however, a striking accumulation ofbiotin in the segmental cavity between the blastoderm and YSLwas found in morphants. Of the morpholino injected embryos,∼50% showed increased biotin permeability, hence correlatingwith the observed penetration rate of 40–60% for the phenotype

based on morphological assessment. A similar increased accu-mulation of biotin was observed in ATGmo injected embryos(Supplemental Data II).

To further corroborate the increased biotin permeability inmorphant embryos, we next determined if these embryos weremore sensitive to the toxicity of CdCl2. Embryos were grown to2 dpf in egg water supplemented with CdCl2 at a concentrationof low (0.5 mM) or high (1 mM) toxicity (Konishi et al., 2006)and mortality rates were determined. A significant increase inthe mortality rate of control embryos was only observed in thepresence of 1 mM CdCl2, but not at the lower concentration(Fig. 8B). In contrast, the mortality rate of tjp3/zo-3 morphantswas already significantly higher at 0.5 mM CdCl2 as comparedto untreated morphants (Fig. 8B; 2-way ANOVA, pb0.01). An

Fig. 8. Tjp3/Zo-3 is essential for EVL permeability barrier function. (A) Biotinpermeability assay: negative control without biotin (a), control morpholino (b)and tjp3/zo-3 morpholino (c, d) injected embryos were dechorionized at 6 hpfand exposed to 1 mg/ml Biotin in egg water. Cryosections were probed withlabeled streptavidin and images taken using identical parameters. In Cmoinjected embryos, although some background staining for streptavidin wasobserved (a), only little biotin diffusion was detected (b). In 40% of the tjp3/zo-3morpholino injected embryos, a significant accumulation of biotin was observedin the segmentation cavity between the blastoderm and the yolk syncytial layer(c, d, arrows), demonstrating a compromised epidermal barrier in embryoslacking Tjp3/Zo-3. (B) Morphant embryos are more sensitive to the toxicheavy metal cadmium as compared to control embryos. Embryos injectedwith either control morpholino (Cmo) or tjp3/zo-3 splice morpholino (MO) weredechorionized at 6 hpf and grown in 1× egg water as a negative control or in eggwater containing 0.5 mMCdCl2 or 1 mMCdCl2. Significance of differences wasestablished using 2-way analysis of variance. Cmo embryos had a mortality of∼20% after 42 h exposure to low toxicity CdCl2, whereas the mortality ofmorphants was significantly higher at ∼50% (pb0.01). At the high cadmiumtoxicity, Cmo embryos had ∼50% mortality and mortality of MO injectedembryos was again significantly higher at ∼80% (pb0.05). Mean and standarderror (S.E.M.); n≥20 per experiment; 3–4 independent experiments. 2-wayANOVA; pb0.01. Asterisks indicate statistical significance.

Fig. 9. Tjp3/zo-3 morphants are sensitive to osmotic stress. (A) The ability oftjp3/zo-3 morphants to survive high salt conditions is significantly reducedcompared to control or rescued embryos. Embryos were either injected withcontrol morpholino or tjp3/zo-3 splice morpholino or they were rescued bycoinjecting MO with pcDNA3.1 containing zebrafish tjp3/zo-3 (MO & zf tjp3).Embryos were dechorionized and exposed to high salt conditions (see Materialsand methods) at 6 hpf. Surviving embryos were counted at 1 dpf. A significantfraction of tjp3/zo-3 morpholino injected embryos died prematurely comparedto control embryos. Mean and S.E.M.; n≥20 per experiment; 3–5 independentexperiments. t-test; pb0.05. Asterisks indicate statistical significance. (B)Selection against tjp3/zo-3 morphants exposed to high salt or high magnesiumduring embryogenesis. High sodium and calcium salt concentrations reduced thesurvival of both normal (no phenotype) and morphant embryos. However, in100× egg water and in 20× MgSO4 the number of morphants that survived to2 dpf was significantly lower than that of normal embryos. Mean and S.E.M.;n≥20 per experiment; 3–5 independent experiments. 2-way ANOVA; pb0.05;statistical significance is indicated by asterisks.


increased permeability to small tracers and enhanced suscept-ibility to heavy metals in morphants is thus consistent with acompromised epidermal barrier function.

Tjp3/zo-3 morphants are sensitive to osmotic stress andosmotic stress mimics the morphant phenotype

Osmoregulation is critical for freshwater fish that live in ahypo-osmotic environment. Besides the kidneys and the gills,the skin plays a critical role in the adult fish in regulating theexchange of ions and water with the environment (Rombough,

2002). Given the importance of TJs in controlling paracellularion permeability and the apparent role of Tjp3/Zo-3 in EVLbarrier function, we explored the effect of osmotic stress ontjp3/zo-3 morphants.

To induce osmotic stress, control or morphant embryos weregrown in increasing concentrations of physiological salts. Theability of MO injected embryos to survive high salt conditionswas significantly reduced compared to either control embryosor embryos coinjected with a zebrafish tjp3/zo-3 cDNA torescue the phenotype (Fig. 9A; t-tests, pb0.05). Importantly,the increased mortality could be directly linked to the tjp3/zo-3MO as high salt and high Mg2+ concentrations selected againstthe morphants (Fig. 9B; 2-way ANOVA, pb0.05).


Due to the hypo-osmotic environment of egg water, a defectin the epidermal barrier function is expected to result in passivewater influx, which, if it exceeds the volume that can be clearedby the kidneys, will accumulate in the pericardium and causeedema. Transfer of such embryos from normal egg water into aniso- or hyper-osmotic sucrose solution to dissipate the osmoticgradient alleviates edema, whereas normal embryos die ofdehydration (Hentschel et al., 2005). Indeed, ∼80% of the tjp3/zo-3morphants present with pericardial edema (Fig. 4A). Whilethe incidence of edema of 2 dpf tjp3/zo-3 morphants selectedfor phenotypes (i.e. edema, curved tail, small tail fin) placed inegg water did not change over a 24 hr period, a significantdecrease was observed in a hyperosmotic sucrose solution (Fig.10A; 2-way ANOVA, pb0.01). The lower incidence of edemawas not due to an increased mortality, as their mortality rate wassimilarly low as that for control embryos until 3 dpf (Fig. 10B).

Intriguingly, we observed that WT embryos grown in highsalt conditions frequently developed abnormalities that showeda striking similarity to the tjp3/zo-3 morphant phenotype, in-cluding a small tail fin and, by 48–56 hpf, defects in circulation

Fig. 10. Compromised osmotic homeostasis mimics the tjp3/zo-3 morphant phenotypembryos in a hyperosmotic sucrose solution. At 2 dpf tjp3/zo-3 morphants were transwith edema (∼20% had a curved tail only). If grown in 1× egg water (diamonds)However, if tjp3/zo-3 morphants were grown in 300 mM Sucrose (triangles), edemaindependent experiments. 2-way ANOVA; pb0.01. (B) Mortality of embryos grownmortality of only ∼20% at 4 dpf (diamonds), whereas the mortality for controls wasmost morpholino injected (triangles) and control (open triangles) embryos survivedclose to 90% of both control and morphant embryos died. There was no statistical sigMean and S.E.M.; n≥10 per experiment; 2–6 independent experiments. Two-way Atjp3/zo-3morphant phenotype. WTembryos in 1× egg water have a straight tail, a widwith curved tail, reduced tailfin (b, arrowheads), the loss of blood circulation and an a100× egg water (c), 20× MgSO4 (d), 10× NaCl or 20× CaCl2 (data not shown), aaccumulation of red blood cells (d, arrow).

leading to the accumulation of red blood cells in the tail region(Fig. 10C). This observation indicates that tipping the osmoticbalance, either due to osmotic stress or a compromised epi-dermal barrier, results in similar phenotypic effects on embryos.

Discussion

In the present study we explored the role of the TJ adaptorprotein Tjp3/Zo-3 in the developing zebrafish embryo using amorpholino knock down approach. The major phenotypes ofthe morphant embryos are edema, loss of blood circulation andtail fin malformations. Silencing of tjp3/zo-3 results in the dis-ruption of the ultrastructure of TJs of the EVL, without affectingcell polarity. Morphants show a loss of epidermal barrierfunction, as assessed by an increased permeability of the EVL tolow molecular tracers and a higher sensitivity of the embryos toosmotic stress.

During mouse development, TJP1/ZO-1 is incorporatedinto sites of cell–cell adhesion mediated by E-cadherin at thecompacted 8-cell embryonic stage (Fleming et al., 2000). TJs

e. (A) Pericardial edema of tjp3/zo-3 morphants can be reversed by growing theferred into a 300 mM sucrose solution. ∼80% of tjp3/zo-3 morphants presented, the incidence of edema only slightly decreased, due to individuals that died.incidence dropped significantly. Mean and S.E.M.; n≥10 per experiment; 5–6in 300 mM sucrose. Morpholino injected embryos grown in 1× egg water had avery low (open diamonds). Placed in a hyperosmotic sucrose solution at 2 dpf,up to 3 dpf, when edema incidence was counted. However, after 48 h exposurenificance between mortality of control or morphant embryos in 300 mM sucrose.NOVA; pN0.5. (C) Exposure of WT embryos to high salt conditions mimics thee tailfin and circulating blood (a). Tjp3/zo-3morphants in 1× egg water presentedccumulation of red blood cells (b, arrow). 10–15% of control embryos raised inlso presented with reduced tailfin (c, d, arrowheads), no blood circulation and


then mature by the incorporation of the TJP1/ZO-1 a+-isoform,cingulin and rab13 and segregation of the TJ components fromthe lateral AJs to establish a permeability barrier around the 32-cell stage (Fleming et al., 1993). In the Xenopus embryo,establishment of the permeability barrier is controversial andhas been reported to occur as early as the 2-cell stage (Fesenkoet al., 2000) or as late as the 64-cell stage (Merzdorf et al.,1998). In the zebrafish, Tjp1/Zo-1 is present in cell–cell junc-tions of the 8-cell stage embryo. The incorporation of TJP3/ZO-3 has not been analyzed in mouse or Xenopus. Tjp3/Zo-3 isfirst detected at the 64-cell stage in zebrafish, consistent with alate role in the functional maturation of TJs.

After implantation of the mouse embryo, TJP1/ZO-1 isrestricted to the trophectoderm and absent from the inner cellmass (Collins et al., 1995). The trophectoderm forms a per-meability barrier between the developing embryo and theuterine environment. In the zebrafish, an analogous function canbe attributed to the EVL, which shields the embryo from theaquatic environment (Kimmel et al., 1990). Besides tjp3/zo-3,which is highly expressed in the EVL, additional TJ com-ponents such as tjp1/zo-1 and tjp2/zo-2 (data not shown) areexpressed in this cell layer. In addition to its localization to sitesof cell–cell contact in the EVL, Tjp3/Zo-3 was also detected inthe nucleus at 6 hpf. The relevance of this nuclear localization isnot clear since, in contrast to the other two Tjps, Tjp3/Zo-3 hasnot been linked to any nuclear activity so far. The presence ofbona fide TJs in the EVL is evident from the presence of typicalelectron dense plaques at the luminal end of the lateral mem-brane of adjoining cells as observed by TEM.

The physiological role of TJP3/ZO-3 has recently beenexplored by inactivating the corresponding gene inmice (Adachiet al., 2006; Xu et al., 2008). Surprisingly, TJP3/ZO-3 knock-outmice lack an apparent phenotype, indicating that in mammalsTJP3/ZO-3 is dispensable. In contrast, silencing of tjp3/zo-3 inzebrafish embryos using morpholinos results in a well-definedand specific phenotype. Zebrafish tjp3/zo-3 but, interestingly,not canine TJP3/ZO-3 rescued the phenotype. This may reflect aunique function of the tjp3/zo-3 splice variant predominantlyexpressed in zebrafish embryos, which differs in its domainorganization from mammalian TJP3/ZO-3 (Kiener et al., 2007).

The structural integrity of TJs in the EVL of morphants iscompromised based on discontinuities and the absence ofelectron dense junctional plaques. Nevertheless, Tjp1/Zo-1localization is not dramatically altered and silencing of tjp3/zo-3 does not affect cell polarity as assessed by the intactasymmetrical distribution of apical and basolateral markerproteins in cells of the EVL and the kidney. Establishing epi-thelial cell polarity requires the Crumbs complex, containingCrumbs, Pals1 and PATJ (Margolis and Borg, 2005). Silencingof PATJ in tissue culture cells results in the basolateral mis-localization of several TJ proteins, including TJP3/ZO-3 andoccludin, a delay in TJ formation and defects in cell polarization(Michel et al., 2005; Roh et al., 2003). The C-terminal PDZ-binding motif in TJP3/ZO-3 interacts with a PDZ domain inPATJ, thereby linking the Crumbs complex to TJs (Roh et al.,2002). The finding that both WT zebrafish Tjp3/Zo-3 and amutant lacking the C-terminal PDZ-binding motif rescue the

morphant phenotypes indicates that these are not due to theinability of Crumbs to link to TJs and is consistent with theintact cell polarity of the morphants. Furthermore, the mutantTjp3/Zo-3 lacking the C-terminal PDZ-binding motif showed asimilar subcellular localization in cells of the EVL and pro-nephric duct as the WT protein and, as assessed by RT-PCR, theloss of Tjp3/Zo-3 was not compensated by overexpression ofcomponents of the Crumbs complex in morphant embryos (datanot shown).

Several lines of evidence revealed defects in the barrierfunction of the EVL of tjp3/zo-3 morphant embryos. Theseinclude the increased permeability of the EVL to small mole-cules and a higher sensitivity of embryos to toxic heavy metalsand osmotic stress. Given the critical role of members of theclaudin protein family in forming the TJ permeability barrierand the role of ZO proteins in linking claudins to the underlyingcytoskeleton (Tsukita and Furuse, 2000; Van Itallie andAnderson, 2006), our data indicates that tjp3/zo-3 in zebrafishis critical for the ability of claudins in the EVL to establish afunctional TJ barrier. At least 25 distinct claudin genes areexpressed in the skin of teleosts (Loh et al., 2004). Due to theimportance for the interaction of Claudin-16 with ZO-1 for itslocalization to TJs (Muller et al., 2003), it would be of interest todetermine if any of the claudins in the EVL of tjp3/zo-3morphants are mislocalized. Unfortunately, such an analysis iscurrently not feasible because of both the large number ofdifferent claudin genes present in teleosts and the lack ofantibodies. In this context, it is of interest to note that, reminiscentof the claudinj zebrafish mutant (Hardison et al., 2005), tjp3/zo-3morphants show a reduction of otholiths (Fig. 3b).

Pericardial edema and arrest of blood circulation, as ob-served in the tjp3/zo-3 morphants, are consistent with defects inosmoregulation and commonly observed if fish are exposed totoxins that impair fluid homeostasis by targeting either thekidney (Hentschel et al., 2005) or the skin (Hill et al., 2004).Edema incidence in tjp3/zo-3 morphant embryos droppedsignificantly when the osmotic gradient that drives water influxinto the embryo was dissipated. Adult fish rely on the kidneys,the gills and the skin for maintaining ion and water balance.Renal excretion is the primary means to eliminate surplus waterbefore the development of functional gills (Rombough, 2002).Prior to the onset of glomerular filtration, which in zebrafishembryos starts as early 48 hpf (Drummond et al., 1998), theEVL barrier likely plays a key role in osmoregulation (Kellerand Trinkaus, 1987).

In conclusion, silencing tjp3/zo-3 leads to tail fin malfor-mations, pericardial edema and arrest of blood circulation inzebrafish embryos. These phenotypes are a consequence ofthe compromised osmoregulation and linked to the loss of thepermeability barrier of the EVL. Thus Tjp3/Zo-3 is a com-ponent of TJs of the EVL of zebrafish embryos and is critical forthe function of the EVL as a diffusion barrier.

Acknowledgments

We thank Jovienne Ee Phei San and Nicole Tsang Ying Hungfor excellent technical assistance, all members of the Zebrafish


scientific community at IMCB, in particular Vladimir Korzhand Steven Haw Tien Fong, as well as the staff of the IMCBZebrafish facility for their support, helpful discussions andexpert assistance. This work was supported by the Agency forScience, Technology and Research (A*STAR), Singapore.

Appendix A. Supplementary data

Supplementary data associated with this article can be found,in the online version, at doi:10.1016/j.ydbio.2007.12.047.

References

Adachi, M., et al., 2006. Normal establishment of epithelial tight junctionsin mice and cultured cells lacking expression of ZO-3, a tight-junctionMAGUK protein. Mol. Cell. Biol. 26, 9003–9015.

Anderson, J.M., et al., 2004. Setting up a selective barrier at the apical junctioncomplex. Curr. Opin. Cell Biol. 16, 140–145.

Collins, J.E., et al., 1995. Regulation of desmocollin transcription in mousepreimplantation embryos. Development 121, 743–753.

Drummond, I.A., et al., 1998. Early development of the zebrafish pronephrosand analysis of mutations affecting pronephric function. Development 125,4655–4667.

Fesenko, I., et al., 2000. Tight junction biogenesis in the early Xenopus embryo.Mech. Dev. 96, 51–65.

Fleming, T.P., et al., 1993. Localisation of tight junction protein cingulin istemporally and spatially regulated during early mouse development.Development 117, 1135–1144.

Fleming, T.P., et al., 2000. Assembly of tight junctions during early vertebratedevelopment. Semin. Cell Dev. Biol. 11, 291–299.

Furuse, M., et al., 2002. Claudin-based tight junctions are crucial for themammalian epidermal barrier: a lesson from claudin-1-deficient mice.J. Cell Biol. 156, 1099–1111.

Gonzalez-Mariscal, L., et al., 2000. MAGUK proteins: structure and role in thetight junction. Semin. Cell Dev. Biol. 11, 315–324.

Gonzalez-Mariscal, L., et al., 2003. Tight junction proteins. Prog. Biophys. Mol.Biol. 81, 1–44.

Hardison, A.L., et al., 2005. The zebrafish gene claudinj is essential for normalear function and important for the formation of the otoliths. Mech. Dev. 122,949–958.

Hauptmann, G., Gerster, T., 1994. Two-color whole-mount in situ hybridizationto vertebrate and Drosophila embryos. Trends Genet. 10, 266.

Hentschel, D.M., et al., 2005. Acute renal failure in zebrafish: a novelsystem to study a complex disease. Am. J. Physiol.: Renal. Physiol. 288,F923–F929.

Hill, A.J., et al., 2004. Water permeability and TCDD-induced edema inzebrafish early-life stages. Toxicol. Sci. 78, 78–87.

Keller, R.E., Trinkaus, J.P., 1987. Rearrangement of enveloping layer cells

without disruption of the epithelial permeability barrier as a factor inFundulus epiboly. Dev. Biol. 120, 12–24.

Kiener, T.K., et al., 2007. Identification, tissue distribution and developmentalexpression of tjp1/zo-1, tjp2/zo-2 and tjp3/zo-3 in the zebrafish, Daniorerio. Gene Expression Patterns 7, 767–776.

Kimmel, C.B., et al., 1990. Origin and organization of the zebrafish fate map.Development 108, 581–594.

Kimmel, C.B., et al., 1995. Stages of embryonic development of the zebrafish.Dev. Dyn. 203, 253–310.

Konishi, T., et al., 2006. Enhancing the tolerance of zebrafish (Danio rerio) toheavy metal toxicity by the expression of plant phytochelatin synthase.J. Biotechnol. 122, 316–325.

Loh, Y., et al., 2004. Extensive expansion of the Claudin gene family in theteleost fish, Fugu rubripes. Genome Res. 14, 1248–1257.

Margolis, B., Borg, J.P., 2005. Apicobasal polarity complexes. J. Cell. Sci. 118,5157–5159.

Matter, K., Balda, M.S., 2007. Epithelial tight junctions, gene expression andnucleo-junctional interplay. J. Cell. Sci. 120, 1505–1511.

McNeil, E., et al., 2006. Zonula occludens-1 function in the assembly of tightjunctions in Madin–Darby canine kidney epithelial cells. Mol. Biol. Cell 17,1922–1932.

Merzdorf, C.S., et al., 1998. Formation of functional tight junctions in Xenopusembryos. Dev. Biol. 195, 187–203.

Michel, D., et al., 2005. PATJ connects and stabilizes apical and lateral com-ponents of tight junctions in human intestinal cells. J. Cell. Sci. 118,4049–4057.

Muller, D., 2003. A novel claudin 16 mutation associated with childhoodhypercalciuria abolishes binding to ZO-1 and results in lysosomal mis-targeting. Am. J. Hum. Genet. 73, 1293–1301.

Roh, M.H., et al., 2002. The carboxyl terminus of zona occludens-3 binds andrecruits a mammalian homologue of discs lost to tight junctions. J. Biol.Chem. 277, 27501–27509.

Roh, M.H., et al., 2003. The Crumbs3–Pals1 complex participates in theestablishment of polarity in mammalian epithelial cells. J. Cell. Sci. 116,2895–2906.

Rombough, P., 2002. Gills are needed for ionoregulation before they are neededfor O(2) uptake in developing zebrafish, Danio rerio. J. Exp. Biol. 205,1787–1794.

Tsukita, S., Furuse, M., 2000. Pores in the wall: claudins constitute tight junctionstrands containing aqueous pores. J. Cell Biol. 149, 13–16.

Tsukita, S., et al., 2001. Multifunctional strands in tight junctions. Nat. Rev.,Mol. Cell Biol. 2, 285–293.

Umeda, K., 2004. Establishment and characterization of cultured epithelial cellslacking expression of ZO-1. J. Biol. Chem. 279, 44785–44794.

Umeda, K., et al., 2006. ZO-1 and ZO-2 independently determine where claudinsare polymerized in tight-junction strand formation. Cell 126, 741–754.

Van Itallie, C.M., Anderson, J.M., 2006. Claudins and epithelial paracellulartransport. Annu. Rev. Physiol. 68, 403–429.

Xu, J., et al., in press. Early embryonic lethality of mice lacking ZO-2, but notZO-3, reveals critical and non-redundant roles for individual ZO proteins inmammalian development. Mol. Cell. Biol. doi:10.1128/MCB.00891-07.

http://dx.doi.org/doi:10.1016/j.ydbio.2007.12.047

tjp3/zo-3 is critical for epidermal barrier function in zebrafish embryos

Documents